v0.9.3

Enhancements 🎉

• Add the ability to run synthetic jobs on CPUs when using gretel agent c3ae1bc
• Use volume-builder image to prepare data volumes during local runs 143fd41

Bugfixes 🐛

• Model id wasn't being prefixed to agent worker logs 143fd41

v0.9.2

Enhancements 🎉

• Updates to configure_session to make it more notebook friendly 072fd03
• Update agent to support custom environment variables and volumes 7045740
• Update agent to support launching on-prem GPU jobs 07be637

Housekeeping 🧹

• black formatting updates

v0.9.1

Bugfixes 🐛

• Worker wouldn't detach from agent fb82d3b
• Project suffix not always unique when using create_or_get_unique_project c0e909f
• Record generation would sometimes fail with default --max-invalid value 048c8b5

Housekeeping 🧹

• Updating billing copy to reflect GA d85106a

v0.9.0

Enhancements 🎉

• Enable uploading Parquet datasets 718b7ee
• [Preview] Add connector related commands to the CLI b9fd267

Bugfixes 🐛

• Fix Docker GPU check 1d49296

Housekeeping 🧹

• Pin transitive Sphinx dependencies 6c868bb
• Introduce more generic support for different workflows in the CLI e6a02ae
  • ⚠️ This update introduced backward incompatible API change: Model.model_type now returns a string instead of enum. The returned string value is the same as value of the previously returned enum.

v0.8.5

Enhancements 🎉

• Add backoff and retry logic for rate limited requests edbe227
• Allow for model name updates from the CLI and SDK a2a431c, 9f38b07
• Improved CLI error handling c36701c
• Add http retry logic when opening and validating remote datasets a1cf537
• Add SDK method for fetching signed project artifact links a393d7f
• Add new UsersApi python bindings b475304
• Add helper method create_or_get_unique_project for creating unique named projects by user b475304
• Add helper method do_api_call for accessing APIs not yet available via python bindings b475304

Bugfixes 🐛

• Docker pull progress wasn't correctly being rendered when using the SDK in a notebook 49eff55

Housekeeping 🧹

• Rely on the API to delete project subresources 7d59444
• Fix failing sphinx builds c1eae89

v0.8.4

Enhancements 🎉

• Show docker pull progress when launching new worker containers 39cff28
• Add privacy_protection_level to synthetics report output 97fecdb
• Update API calls when fetching worker logs c993dc8
• Improved retry logic on transient HTTP errors c6b54ea

Bugfixes 🐛

• Fix record handler output when running jobs in manual mode 02009e5
• Unable to download run reports due to an incorrectly configured artifact key ddb51c3

Housekeeping 🧹

• Format and lint codebase 67c29a4

v0.8.3

Enhancements 🎉

• gretel projects search API exception handling is more consistent.

Bugfixes 🐛

• submit_docker_local now correctly configures a GPU when training synthetic models.

v0.8.2

Enhancements 🎉

• New submit_cloud and submit_manual Job methods make it easier to create and submit a model or record handler. These methods will replace the existing submit method in future releases.
• Project.search_models now returns a Model instance instead of a dict.
• Added a new helper method, submit_docker_local that makes it easier to launch local Jobs via docker.

Bugfixes 🐛

• CLI flags and environment variables now correctly take precedence over configuration file settings.
• Cloud models no longer are downloaded from the CLI after training.
• Local docker runs now gracefully fail when docker isn't running or installed on the host machine.

v0.8.1: Bugfixes

• Bugfix: API requests would fail SSL cert verification on certain machine configurations.

Support for Beta2 APIs and Workflows

This release supports Gretel's Beta2 workflows and APIs 🥳.

For more information about the release, please refer to this blog post, https://gretel.ai/blog/beta2-privacy-engineering-as-a-service.

Note: The 0.8.0 release contains a number of breaking changes from 0.7.0.